Enabling first-principles predictions of core performance in tokamaks with surrogate optimization techniques

Pablo Rodriguez-Fernandez

MIT PSFC

Friday, March 3, 2023

3:00pm

NW17-218 Hybrid

PSFC Seminars

Abstract: Predicting core gradients in tokamak plasmas using first-principles based models has long been the goal of transport research, but it has largely been intractable due to the large computational expense. Given this high computational cost, the transport community has relied on a set of tools that span from empirical methods (POPCON) to quasilinear models of turbulence (integrated modeling toolsets) to scope future devices, while only using full nonlinear gyrokinetic simulations as standalone, spot-check validation studies. This talk will present a newly developed flux-matching technique that reduces the computational cost of first-principles, multi-channel, non-linear gyrokinetic predictions of the plasma core by, at least, a factor of 5 without compromising accuracy. Thanks to this, and the performance improvements of the CGYRO code developed by collaborators at General Atomics, core profile predictions from first principles are now able to be routinely performed on modern clusters and supercomputers. This talk will introduce the need of flux-driven simulations to enable profile predictions, will present the fundamentals of the novel PORTALS technique and will discuss recent applications: the study of performance of burning plasmas in ITER and SPARC and the validation of ion-scale gyrokinetics in DIII-D and JET experiments.